New forensic evidence suggests that the voice behind ChatGPT's 'Sky' assistant bears a striking resemblance to actress Scarlett Johansson, a finding that could bolster her legal position against OpenAI. Researchers at Arizona State University, in a study commissioned by NPR, compared Sky's voice to those of approximately 600 actresses using multiple AI models. The results placed Johansson's voice closer to Sky than 98% of the other candidates.
The analysis, led by computer scientist Visar Berisha, used advanced voice-matching algorithms to evaluate acoustic features. While the study found that Johansson's voice was not always the top match—with actresses Anne Hathaway and Keri Russell occasionally ranking higher—the overall similarity was statistically significant. Berisha told NPR, 'Our analysis shows that the two voices are similar but likely not identical.'
One particularly notable finding involved vocal tract length, a physical characteristic that shapes voice. The simulation indicated that Sky and Johansson would have identical vocal tract lengths, a rare match that suggests a fundamental acoustic alignment. However, the study also noted differences: Sky's voice is slightly higher pitched and more expressive, while Johansson's is breathier.
Background of the Controversy
The controversy began last month when OpenAI released a major update to ChatGPT, introducing a new voice assistant capable of real-time conversation. Sky was one of the available voices, and users quickly noticed its resemblance to Johansson's performance in the 2013 film 'Her,' where she voices an AI chatbot. The comparison was fueled by OpenAI CEO Sam Altman, a known fan of the movie, who posted the single word 'her' on social media the day the voice assistant launched.
Days later, OpenAI abruptly removed the Sky voice, claiming it had not copied Johansson. The company said a different actress was behind the voice, a claim later corroborated by The Washington Post. Johansson then revealed that OpenAI had approached her twice to license her voice, and she had declined both offers. She expressed shock at the similarity, calling it 'eerily similar' to her own.
Legal Implications
OpenAI has attributed the incident to poor communication rather than intentional copying. Johansson has not filed a lawsuit but has retained legal counsel. Legal experts have suggested she could have a strong case, and the new forensic findings may strengthen her position. The study's methodology and results have not been peer-reviewed, but they add a scientific dimension to a dispute that has drawn widespread attention to AI ethics and voice rights.
As the debate over AI-generated voices continues, this case highlights the challenges of protecting personal identity in an era of advanced synthetic media. The outcome could set a precedent for how voice actors and celebrities are compensated and credited in AI applications.
Comments
Sign in to leave a comment
No account? Create one
No comments yet.